LLVM/Clang 10.0 Adds AMD Zen 2 Scheduler Model For Optimized Code Generation

[phoronix]

Phoronix: LLVM/Clang 10.0 Adds AMD Zen 2 Scheduler Model For Optimized Code Generation

It's too bad that it has taken so many months after AMD Zen 2 based Ryzen and EPYC processors began shipping to see this compiler support in place, but the good news now is that for the upcoming release of LLVM 10.0 is now the Zen 2 scheduler model being added to the "znver2" target...

http://www.phoronix.com/scan.php?page=news_item&px=LLVM-10-AMD-Znver2-Sched-Model

[ms178]

Maybe the new HPC design wins will put some pressure on AMD to get better support out sooner. There are other areas (e.g. Glibc) where they still need to engage with the community to provide the best performance for their products.

[letrout]

With this, does stock LLVM finally supersede the AOCC? Or does AOCC still contain other optimizations that have not made it upstream?

[Marc Driftmeyer]

Maybe the new HPC design wins will put some pressure on AMD to get better support out sooner. There are other areas (e.g. Glibc) where they still need to engage with the community to provide the best performance for their products.

Has it occurred to anyone that presently AMD resources are engaged directly with major clients optimizing their stacks for their needs first? Enterprise clients come first and always get early access before any upstreaming occurs.

[ms178]

Has it occurred to anyone that presently AMD resources are engaged directly with major clients optimizing their stacks for their needs first? Enterprise clients come first and always get early access before any upstreaming occurs.

Right, but I guess they are already doing this work since they had their first Naples engineering samples ready and nonetheless, the Glibc issue I pointed out here specifically is still not addressed. I'd consider this to be a top priority. Just look at the ARM vendors which also contribute to Glibc long before hardware is available on the market.

[Royi]

They need to invest in some optimized BLAS library that can rival the performance of Intel MKL.